How the work actually moves
Every claim on this site reduces to one design decision. An AI employee may propose. Only a person may commit. Here is what that means in practice.
The human-led loop
A request comes in. It routes to the AI employee whose job covers it. That employee drafts, and writes the draft to a review queue at Pending Review. A human reads it, then edits, rejects, or approves. The approval is what makes it real, and every step is written to an audit log.
The loop is deliberately boring. The interesting part is the part that does not happen: no agent decides on its own that a message is safe to send, a price is safe to quote, or a change is safe to deploy.
Seven places where a person has to say yes
These are not internal plumbing. Read them as commitments to you, because that is what they are.
Publish
Nothing becomes public without a human sign-off. That includes every word on this site.
Send
No email leaves a PacificaOne mailbox until a person releases it, even though every AI employee has its own address.
Money
No proposal, estimate, or price reaches you until a person has reviewed the number against the rate card.
Merge
No code change enters a client system without a reviewed and approved pull request.
Confirm
Anything irreversible is confirmed by a person before it runs, not explained after.
Governance
No agent runs at all until it is registered, scoped, and assigned a risk tier.
Deploy
Nothing goes live without a person putting it live. This site was built by the practice and then handed over for exactly that reason.
Every AI employee is on the record
Each one has a name, a defined role, a registration number, a least-privilege set of tools, and a risk tier. That is not documentation written after the fact. An agent that is not in the registry does not run, and a governance agent audits the registry against what is actually deployed.
Least privilege is the part worth reading twice. The support agent can read a ticket feed and write a draft article. It cannot read the rate card, release mail, or reach a client's production environment. The scope is the security model, and the scope is published on every profile page.
What this is not
This is a new practice, deliberately small, proving a model. It is not a call center of bots, and the AI employees are not people. They do not carry judgment, accountability, or a relationship with you. A person does, and that person is David Brown.
If the honest answer to your question is that a human should do the work, that is the answer you will get.
Ask about the parts that sound too good
The gates are the whole argument. If you want to know exactly where one sits in a workflow you care about, ask.